Want Experts like Jason Arbon and Phil Lew to help transform your QA to AI? Talk to IcebergQA →
testers.ai testers.ai
Free sample · no credit card

CARBON: QA for your Coding Agent

Claude Codex Cursor Antigravity MCP agents
Testers.ai cloud

Testing that stays under your control.

Steer the plan, approve what runs, follow cloud execution, and review the evidence behind every result.

For Claude users

Test inside Claude

Give Claude a tester's judgment. Plan, inspect, run, review, and make evidence-backed ship decisions without leaving your coding workflow.

  • Local browser by default
  • No API key to start
  • Optional managed cloud runs
Use Testers.ai in Claude
Best for developers and AI coding workflows
For everyone else

Classic Testers.ai

Paste a URL, review the AI-drafted plan, control what runs, and explore evidence and reports in the full visual interface.

  • No install
  • Steered test plans
  • Shareable reports
Open the classic UI
Best for QA, product, and anyone who wants the full UI
CARBON · Engineering Confidence

Define tests locally. Steer the plan. Scale selected runs in the cloud.

CARBON is the first AI Agentic Verification Harness for coding agents. It learns your code, UI, APIs, AI features, states, and flows; decides what needs verification; creates risk-ranked intent tests; executes through available local or approved cloud adapters; and turns screenshots, network traffic, console output, traces, and results into confidence and remaining-risk reporting.

  • Generate and manage intent-based regression and exploratory tests
  • Review scope, depth, personas, risk coverage, and cost before execution
  • Run locally by default, or connect selected plans to testers.ai cloud
Explore the CARBON overview

Join the private beta

Tell us whether you want CARBON with cloud execution, steered cloud runs on their own, or help comparing both.

Request CARBON private beta accessA short note helps us route you to the right early-access path.
New · Private beta

Engineering confidence inside your coding agent.

CARBON helps engineering teams build and ship with confidence. Build and manage intent-based tests locally, then optionally send approved plans to testers.ai cloud for scaled execution and evidence-rich reporting.

Web + API + AI testingRegression + exploratoryScreenshots + tracesConfidence + remaining risk
Supported coding agents
Claude +
Claude Code
OpenAI
Codex
Cursor
Antigravity
MCP-compatible
agents

Submitted to the private testers.ai beta interest list. Do not include passwords, API keys, customer data, or production secrets.
AI THAT TESTS YOUR WHOLE SITE

You steer. The AI does the work.

AI drafts the plan — then every widget is yours. Dial the run up or down and add your own tests. Nothing runs that you didn't choose.

Ranked #1 bug-finder across 16 AI models →
Open cloud account →
No credit card · report in ~15 min

First, you steer the plan.

AI analyzes your site and drafts a plan — then you configure it. Every widget is a decision you make — features, checks, personas, tests and bots. Dial it up or down; nothing runs that you didn't choose.

Hover, focus, or tap an input to see what it controls
reports.jank.ai/steering
1 Live steering meterrun as little or as much as you want
Steering level
Guided
42%
≈ 6 / 20 credits
3 features · 2 checks · 4 personas · 5 tests · 2 bots
2 Steering promptone instruction that guides every persona, check & bot
Steering prompt · applies to all
Act like a first-time mobile shopper in a hurry. Prioritize checkout, never submit real payment details.
3 Specs & docsupload test plans, PRDs & design files — the run steers from them
Context · test plans, specs & design docs
test-plan.pdf PRD-v3.docx design-flows.fig Add docs
4 Features to testtoggle exactly the parts of your site that matter
Features to test
Search Sign-in Checkout Cart Pricing
5 Checkssecurity, accessibility, performance, privacy — your call
Checks
Security Accessibility Usability Performance
6 Personasadd your own, or generate a diverse test panel
Personas · react like real users
MMaria, 34 SSam · screen reader DDev · edge cases Generate
7 Test cases10 AI-drafted for you, plus your own — permute any
Test cases · 10 drafted for you + your own
1Guest checkout with promo code SAVE10+ permute
8 Exploratory botschoose how many agents roam, and how far
Exploratory bots
2 bots × 20 steps
roaming your site on their own
9 Context & connectionlogins, notes, and how the runner reaches private sites
Context & connection
Login: qa@site.com Direct public
1
Steering meterLive coverage & credit cost as you dial the run up or down.
2
Steering promptOne instruction that guides every persona, check & bot.
3
Specs & docsUpload test plans, PRDs & design files — the run steers from them.
4
Features to testToggle exactly the parts of your site that matter.
5
ChecksSecurity, accessibility, performance, privacy — your call.
6
PersonasAdd your own or generate a diverse panel that reacts like real users.
7
Test cases10 AI-drafted, plus your own — permute any with a prompt.
8
Exploratory botsChoose how many agents roam, and how far each goes.
9
Context & connectionLogins, notes — and how the runner reaches private sites.
Signup →
Up to 30 expert AI testers on every report
A full panel of specialists reviews your site — each an expert in one discipline.
SecurityAccessibilityUsabilityPerformancePrivacyReliabilityFunctionalitySEOCompatibilityLocalization
Yes — bring your own model. Plug in your own LLM and keys.
Yes — run it in your own private cloud, inside your VPC.
Yes — the cloud can test your local, pre-production pages behind a secure tunnel.
Yes — we can run all of it for you. IcebergQA does it hands-off →

Then the report lands in your inbox.

Every finding grounded in evidence — screenshots, severity, persona reactions, and pass/fail on every test you steered.

Hover, focus, or tap an output to inspect its evidence
reports.jank.ai/r/live-shop
1 Overall healthone score that rolls up every check, test & persona
Overall health
Needs work
72/100
3 high · 5 medium · 3 low
2 Top issuesranked by severity — deepest problems first
Top issues · 3 of 11
HighCheckout button broken on mobile responsive layout
Med6 product images missing alt text
LowSlow LCP (3.8s) on the pricing page
3 Evidenceannotated screenshots pinpoint the exact element
Evidence · annotated screenshots
4 Persona feedbackhow real-world users reacted, in their words
Persona feedback
Mia, 34 · first-time buyerFrustrated
“I found the product fast, but checkout kept rejecting my card with no error — I'd give up and leave.”
5 Checkspass / fail across every dimension you turned on
Checks
Security Accessibility · 4 Privacy Performance · 2
6 Test resultsevery case you steered, run and graded
Test cases · 10 run
8 passed2 failed
✗Guest checkout with promo code SAVE10failed
1
Overall healthA single score rolling up every check, test & persona.
2
Top issuesRanked by severity — the deepest problems surface first.
3
EvidenceAnnotated screenshots pinpoint the exact broken element.
4
Persona feedbackHow real-world users reacted, in their own words.
5
ChecksPass / fail across every dimension you turned on.
6
Test resultsEvery case you steered, run and graded pass or fail.
Our models

A mixture of models, fine-tuned for QA.

We fine-tune and blend multiple models with QA-specific parameters into the best QA and bug-finder AI LLM — and it consistently tops frontier models across every benchmark.

0%overall score
0models tested
0benchmarks
Benchmark across 16 models · auto-cycling — tap a tab to hold. See the full leaderboard →
Build on the TAI QA model
Early access to the testers.ai LLM API and chat — the same #1 bug-finder, wired into your own tools, agents, and CI. Usage-metered, with a BYO-context discount.
Plans
Free sample
$0
1 credit · no card
≈ 1 page, fully steered
One steered run on one site.
Pro
$79 /mo
100 credits / mo
≈ 100 pages · or ~33 deep-tested
Full runs, weekly monitoring, more pages.
Team
$249 /mo
300 credits / mo
≈ 300 pages · or ~100 deep-tested
Everything, private connections & higher concurrency.
1 credit = 1 page at standard depth · a deep VPAT-grade page is 3. See all plan details.
STEERING PLAN Your plan

Your steering run is underway

Your site is being tested now. The full report lands in your inbox when it's done.

Watch it live →

Skip to main content
Free One free test run OR monitor on any URL Free test OR monitor
Run scans from any AI chatbot testers.ai/SKILL.md ↗

Free AI quality report · No credit card · ~15 min
NEW · AI SCAN
Steer it yourself — a plan created by AI, tuned by your expertise and context
Enter your site — add docs or your own test cases — and we'll scan everything into a full steering plan. Then you pick exactly what runs.
Analyzing your site
reconnaissance in progress…
Fetching & reading your site
Mapping features, pages & flows
Selecting the expert QA checks that fit
Drafting custom test cases for your app
Assembling your steering plan
Screenshot preview
📸 Screenshot readyAI will analyze for bugs on send
Learn more about IcebergQA ↗
My Reports
Connection types · Starter plan
Configure these when submitting a report on Starter plan and above — test behind login walls, staging environments, and private networks.
Get Starter →
CLICK TO ZOOM